A Particle Swarm Optimization Framework for Calibration of Driver Behavior Models at a Work Zone

Traffic impact assessment is a key step in the process of work zone planning and scheduling for transportation agencies. Microscopic traffic simulation models enable transportation agencies to conduct detailed analyses of work zone mobility performance measures during the planning and scheduling process. However, traffic simulation results are valid only when the simulation model is calibrated to replicate driver behavior observed in the field. Few studies have provided guidance on the calibration of traffic simulation models at work zones and have offered driver behavior parameters that reproduce capacity values observed in the field. This paper contributes to the existing body of knowledge on work zone traffic simulation by providing separate driver behavior models for heavy vehicles and passenger vehicles at work zones. The driver behavior models are obtained from field data collected at a work zone on Interstate 44 in St. Louis, Missouri. The driver behavior parameters are derived to replicate the flow and speed at the work zone taper and at roadway segments upstream of the work zone. A particle swarm optimization (PSO) framework is proposed to improve the efficiency of the calibration process. The desired time headway was found to be 2.50 seconds for heavy vehicles and 1.89 seconds for passenger cars. The longitudinal following threshold was found to be 39.36 feet for heavy vehicles and 13.12 feet for passenger cars. The proposed driver behavior parameters were tested against field data that had not previously been used in the calibration of driver behavior models. The average absolute relative error (AARE) for volumes at the taper was 7.2% and the GEH index for volumes at the taper was 3.5.